CHEN, Mme Roxane (2024) Semantic segmentation of LiDAR point clouds for power grid power grid infrastructures PFE - Project Graduation, ENSTA.

This is the latest version of this item.

[img]
Preview
PDF
5Mb

Abstract

The LATINA project, initiated by RTE in 2015, aims to monitor high-voltage power lines and surrounding vegetation using airborne LiDAR data and a semantic segmentation model. Despite its initial promise, the project has encountered significant challenges since its production launch in 2021. Considering the progress made in this field and the increasing quantity of data at hand, it was time for a reevaluation of the state of the art methods. This report details the steps taken during my internship within RTE’s software development department. The main objective was to leverage RTE's extensive LiDAR data and train an internal, semantic segmentation model. The focus of the work includes creating a dataset and the evaluation of three models : RandLA-Net, Superpoint Transformer and KP-FCNN.

Item Type:Thesis (PFE - Project Graduation)
Uncontrolled Keywords:3D point-cloud, scene semantic segmentation, machine learning, LiDAR
Subjects:Information and Communication Sciences and Technologies
ID Code:10448
Deposited By:Roxane Chen
Deposited On:28 oct. 2024 12:34
Dernière modification:28 oct. 2024 12:34

Available Versions of this Item

  • Semantic segmentation of LiDAR point clouds for power grid power grid infrastructures (deposited 28 oct. 2024 12:34) [Currently Displayed]

Repository Staff Only: item control page